• Abstract and Figures
• Public Full-text

Final Report: Ecology of the Western Chicken Turtle (Deirochelys reticularia miaria) in the Arkansas Valley: Development of Survey and Monitoring Protocols for a Rare and Secretive Species June 2009 Dr. Stephen Dinkelacker and Nathanael Hilzinger Department of Biology University of Central Arkansas Conway, AR 72035 Introduction: Because of its scarcity and poorly understood biology and ecology, the Western Chicken Turtle (Deirochelys reticularia miaria) has been designated as an S3 species (i.e., rare to uncommon) with fewer than 25 locality records in Arkansas. At this time, development of a conservation and management plan is inconceivable due to a lack of baseline ecological data. This information is critical for the development of suitable conservation strategies and to define appropriate distribution and abundance survey protocols for this rare species. Virtually nothing is known about the Western Chicken Turtle (D. r. miaria) in Arkansas, except that they occur in the state. In fact, little is known about the ecology of the three Chicken Turtle subspecies. Few studies have examined the ecology of Chicken Turtles since 1969, and those were conducted on the eastern subspecies (D. r. reticularia) in Virginia and South Carolina. Most accounts of Chicken Turtles that appear in general herpetological surveys and reference books lack or provide insufficient ecological data. The limited ecological and biological data that are available for Chicken Turtles suggest they use a variety of habitats, including terrestrial habitats, during their annual cycle. Chicken Turtles may have bimodal nesting seasons, with the first season occurring in early spring and the second in late summer/early fall. Preliminary data from a local population suggests that female Western Chicken Turtles are gravid in mid‐June, potentially using terrestrial habitats for most of the year, and occupying seasonal wetlands during late spring and early summer. Survey protocols for Western Chicken Turtles do not exist. In addition, a potentially limited activity season, low population densities, and specific habitat requirements may make the detection and determination of abundance very difficult for this species. In other words, Arkansas currently has no indication of the population status, distribution, or abundance of this species within the state. Because Chicken Turtles potentially occupy small, seasonal wetlands, there is the possibility that Chicken Turtles exist in metapopulations. In other words, individuals may move between small wetlands (e.g., small populations) on a seasonal or potentially yearly basis. Because these seasonal wetlands are subject to environmental fluctuations, as well as anthropogenic impacts, a Chicken Turtle population may be dependant upon the ability to colonize new and re‐colonize old wetlands. The principal goal of this study is to determine and document population structure and size, reproductive ecology, and seasonal habitat selection of Western Chicken Turtles in central Arkansas. We proposed to elucidate these variables and develop appropriate survey protocols including appropriate survey seasons and techniques, which will allow for distribution and abundance surveys to be conducted accurately across the state and throughout the species range. Although official funding for this project began in August 2007, data were collected beginning in 2005. Materials and Methods: Study Site. The study was conducted at a floodplain wetland on the East Fork Cadron Creek located near Holland, Arkansas. The wetland consisted mostly of shallow water, less than 1.5 m in depth, dominated by emergent vegetation including Smartweed (Polygonum spp.), Water‐ primrose (Ludwigia spp.), and Buttonbush (Cephalanthus occidentalis). The surrounding land consisted of a mixed hardwood forest on the eastern side of the wetland and a pine plantation on the western side with Bald‐cypress (Taxodium distichum) located along the water fringes. The wetland was inundated from fall and spring rains with maximum levels obtained during spring when the East Fork Cadron Creek overflowed its banks. After spring rains ceased, the wetland would often dry completely by midsummer until rains began again in the fall (Figures 1‐3). General Methodology. This study was conducted under the Arkansas Game and Fish Commission Scientific Collection Permit 022520083 and in accordance with Institutional Animal Care and Use Committee animal use protocol 2005‐001. Turtles were collected each year (2006 ‐ 2008) from late March through early July using unbaited fyke nets (Vogt, 1980a). A fyke net consists of 2 turtle traps (3 3‐ft rings, 1 inch mesh, Arkansas throat) connected by a 25‐ or 50‐ft lead net (6 ft deep, 1 inch mesh). The traps were submerged at least past the throat, but shallow enough to not be completely submerged (allows turtles to breathe). The concept of the trap is that turtles would swim into the lead net and not be able to swim above or below it. They would then swim or crawl along the net and end up in either trap. Nets were positioned throughout the wetland and checked every 1 ‐ 2 days for turtles. All D. r. miaria were taken to the University of Central Arkansas for data collection. Each turtle was given an individual identification mark by notching a unique pattern into their marginal scutes. Straight‐line carapace length (CL, ± 0.1 mm) was measured for each turtle and sex was determined by the preanal length of a turtle’s tail (Gibbons, 1969). Age was estimated by counting the number of growth rings on the abdominal plastral scutes. However, larger individuals tended to have smooth shells preventing an estimation of their age. Females were examined for eggs through palpation and ultrasound. Two clutches were collected in 2006 by injecting two different gravid females with oxytocin and a third was obtained when a female prematurely expelled her clutch (she flipped herself in a holding pen and was trapped inverted while held in captivity in 2008). Mean clutch size was determined for the three clutches; however, only the two intentionally collected clutches were used to determine mean egg length (± 0.1 mm), width (± 0.1 mm), and mass (± 0.1 g). Eggs were subsequently incubated at 29 °C. Once the eggs hatched, mass (± 0.1 g), CL (± 0.1 mm), and plastron length (PL) (± 0.1 mm) were recorded for each hatchling. The hatchlings were released into the population in spring 2007. Growth models were constructed by using the von Bertalanffy growth model (Fabens, 1965): ‐kt Lt = a (1 ‐ be ), (1) where Lt is CL at age t, a is the maximum CL, b is a variable related to hatchling size, e is the base of the natural logarithm, and k is the intrinsic growth factor. The growth interval method (Fabens, 1965), using the rearranged form of the von Bertalanffy growth model: ‐kd Lr = a ‐ (a ‐ Li)e , (2) where Lr is the CL at recapture, Li is the length at initial capture, and d is the length of time between captures, was used to calculate k for males and females. Carapace length from individuals that were captured at least twice (11 males and 5 females) were used to construct the growth model. The estimated values of k for males and females were then used to solve Equation 1 for b for each sex. Carapace lengths for the growth curves were estimated for males and females (ranging in age from 0 ‐ 25 years) using the estimated values of k and b. Growth rates were compared between the sexes using the estimated CL via an ANCOVA with age as the covariate. Population Modeling. Encounter histories were constructed for each individual, using the method of Cooch and White (2000). During a trapping season (i.e., 2006, 2007, 2008), a turtle was recorded as “captured” if it was caught or as “not captured” if it was never caught during that year. Once a turtle was captured in a season, subsequent recaptures were not considered in the analysis. Four adult turtles were recovered dead during the study, two presumably from predation and two from disease or injury, though the exact causes are uncertain. Capture histories for these four turtles were constructed to represent the dead recoveries as specified by Cooch and White (2000). Turtles were grouped as male, female, or juvenile (i.e., those too small to determine sex with certainty). Population parameters were estimated based on encounter histories using Program MARK (White and Burnham, 1999). Models for apparent survival (Φ) and probability of capture (p) were constructed using the Cormack‐Jolly‐Seber model for open populations (Cormack, 1964; Jolly, 1965; Seber, 1965). A saturated model, which held Φ and p to be dependent on group (i.e., male, female, or juvenile) and time, was fit to the Cormack‐Jolly‐Seber model followed by constructing models with reduced parameters. The fitness of the saturated model was determined by comparing its model deviance to the mean deviance of 1000 bootstrap replicated models. The saturated model deviance approached a significant difference (P = 0.10) from the mean bootstrap deviance sufficiently enough to be taken into consideration during model selection. In order to determine which model best represented the data, model selection was conducted using maximum‐likelihood methods (Burnham et al., 1994). The corrected Akaike Information Criterion (AICc) was used to determine the model that was the most parsimonious (Anderson and Burnham, 1994; Burnham et al., 1995). Due to the degree of overdispersion in the data, as indicated by the bootstrap analysis result approaching a significant P‐value, the model deviance of the AICc (ĉ = 1) was adjusted to the deviance of the saturated model (ĉ = 2.148). The adjusted ĉ required that model selection be conducted using the quasi‐likelihood estimator, QAICc (Burnham and Anderson, 1998). Apparent survival and probability of capture were estimated using weighted averages from the QAICc model weights. Program MARK had difficulty in estimating parameters for some models, resulting in unrealistically high estimates of parameters and standard errors, which affected the weighted averages.

Load more

  29

Copy Link